Comment 1 for bug 1991621

I wonder if this gets better by setting the parameter aligned=True?

def create_memmap_backed_data(data, mmap_mode="r", return_folder=False, aligned=False):

https://sources.debian.org/src/scikit-learn/1.1.2+dfsg-6/sklearn/utils/_testing.py/?hl=639#L639

Maybe just change this *here*; False feels like a *very* silly default. x86 can cope with unaligned by going VERY SLOW. arm64 when emulating armhf tends to just SIGBUS and tell you not to do that. I don't know any other architectures, but unaligned is usually bad news.

Thanks